bulat.ziganshin: > Hello Donald, > > Wednesday, June 27, 2007, 6:37:07 AM, you wrote: > > >> I also know Bulat Ziganshin had put together a nice-looking Streams > >> library (http://unix.freshmeat.net/projects/streams/) based on John > >> Goerzen's previous HVIO work, but I wasn't sure if the ByteString > >> stuff matches the speed and encapsulates all of the functionality of > >> that anyway. Or can/should they be used together somehow? > > > Should be similar in speed, and most high-perf stuff seems to use > > ByteStrings now. ByteStrings also have some nice high level > > optimisations not available to lower level libraries. > > i recommend you to try ByteString own i/o capabilities at first. if it > will not satisfy you, you can try Streams 0.2 which supports > ByteString i/o: > > http://www.haskell.org/library/StreamsBeta.tar.gz > > using my library should allow 30-50 mb/s i/o speed but its > installation may be tricky since it was not updated over a year
That's interesting, Bulat. Two points I'd like to ask about the streams library: What machine did you do the IO benchmarks on? Since we get well over 10x that speed word writing in Data.Binary now, for example, on a fast machine. (Duncan, what's the max throughput we've seen?) And secondly, will the streams stuff be updated? I had trouble benchmarking against it back in January, and am wondering if it will be maintained? -- Don _______________________________________________ Haskell-Cafe mailing list [email protected] http://www.haskell.org/mailman/listinfo/haskell-cafe
